I everything goes smooth I ought to have the last workunits out by tomorrow, or latest the day after.

No problem. We are ready when you are. But the term "the last" is ominous. Is it the end of the road for QC, or just for this experiment?

No, I think there will be more in the future. But we will have to consider well what to run since the large molecules were a bit of a pain. Some molecule fragmentation might solve the issue. But there might be some downtime on QM jobs, hard to tell.

That is perfectly OK. It is much better for projects to tell us what is going on than to leave us hanging. You never know if they are dead or alive. At least you are alive.

So, seems like it will take 2 more days or so since other people are running stuff on the machine that calculates the workunits. It's a bit of a massive amount of data so bear with it a few more days

which machine do you use for calculating the WU? Just in case some of us are willing to donate some money for hardware upgrades at your institute.

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.

Although I love the support, I feel like it's an exercise in futility. Eventually all machines we have become overrun by one or more of our lab members. No machine is safe :D Even if I had one dedicated to myself I might use it for something else while not calculating WUs (would be a shame to have it idle) and then when we run out of WUs I would need to deal with the same issue. I guess we'll just have to wait one or two more days. It's roughly 80% done. Worst case if it's still not done in two days I'll talk with the lab member currently squeezing the living juices out of it to stop his jobs for a moment if possible.

Does QC still have a problem when multiple WUs start at the same time???

I loaded Rosetta jobs and watched them in System Monitor. They appeared to load one by one. The first CPU ramped up to 100% and then a minute or so later the second ramped up... I wonder if they have a solution they'd share.

I'm thinking of a script that could replicate that behavior.
QC WUs DL. The app_config.xml file could be iteratively modified:

<app>
    <name>QC</name>
    <max_concurrent>1</max_concurrent>
</app>

Wait a minute and change 1 to 2 and tell BOINC to Read Config files.

<app>
    <name>QC</name>
    <max_concurrent>2</max_concurrent>
</app>

Wait a minute and change 2 to 3 and tell BOINC to Read Config files.
Repeat until max desired WUs is reached.
Reset N to 1 in case of a reboot you restart one at a time.

E.g., I have a Xeon E5-2699 v4 with 22c/44t. I could conceivably run ten 4-CPU WUs and leave 4 threads for the GPUs.

When the next batch of QC WUs posts I'm going to babysit a computer and do this manually. I've never written a script so I'm reading book learning it now. If someone knows how to code this I'd be glad to test drive their script for them.

Aurum asked:

Does QC still have a problem when multiple WUs start at the same time???

Short answer is No. This was fixed several weeks ago.

Great. How much RAM do I need per QC WU???
Rosetta needs between 750 MB to 1 GB.

Is there an L3 Cache requirement for QC WUs???

E.g. at WCG the MIP project require 4-5s MB per WU. It runs with less but speed drops fast if overloaded.

Today I'll upload a new batch. I have no real clue on the technical requirements though. It will be try-and-see because these are some last large molecules I need to run. After these large ones I'll send some last small and fast ones again.

Wow! This is a major RAM hog. I just watched while 6 4C WUs consumed 16 GB and crashed a computer. Then another had 3 4C WUs running and it stopped one them as Suspended:Wating for memory, then a second one and third is still using 12 of 16 GB. Then they all 3 come back on and use all the memory. Then start failing for computation error.

We may need to reserve 3 GB per C, i.e. run only one 4C WU per 16 GB RAM.
So I'm adding this to my app_config:

<app>
    <name>QC_beta</name>
    <max_concurrent>1</max_concurrent>
</app>
<app_version>
    <app_name>QC_beta</app_name>
    <plan_class>mt</plan_class>
    <avg_ncpus>4</avg_ncpus>
    <cmdline>--fetch_minimal_work</cmdline>
</app_version>

I got two of them. One took 2.5 GB, and the other 3.5 GB. Send more. I need to use the 32 GB of memory on my i7-4770.
And longer ones are OK; these took only 12.75 minutes. I have no idea about disk usage though.

Still beta WU for now. Disk usage may be large (and we can't tune it). Memory should be <4 GB per WU. Threads, up to 4, controlled by boinc. There will be more. Thanks!

I have some TONI WUs that have very low CPU usage and very low RAM usage.

Another thing I noticed is my GPU usage, normally at 100% running GPUGrid is now 5-15% and the CPU is not anywhere near 100% usage.

Update:

I just suspended and resumed the GPU WU and now it's back at 100% usage

Today I'll upload a new batch. I have no real clue on the technical requirements though. It will be try-and-see because these are some last large molecules I need to run. After these large ones I'll send some last small and fast ones again.

No joy with my computers in getting any work. Since they are remote I can't access them to request new work units. Shouldn't be an issue with 1 machine as it has 64GB ram and 4 TB hard drive. The other has 32 GB but also a 4 TB HDD. Just waiting to see if they get any work units to crunch.

---

No joy with my computers in getting any work.

Do you have "Run test applications?" selected in your preferences? I am getting them fairly regularly now.

Can you set it so that each Host gets one beta before a single Host gets 87 ???

No joy with my computers in getting any work.

Do you have "Run test applications?" selected in your preferences? I am getting them fairly regularly now.

Had the beta but not the "Run test applications" selected. Thanks for that. Made the change, will see if that works.

---